ROADEF 2026>
A Genetic Algorithm For The Daily Wildfire Preventive Patrol Routing Problem
Lise Nakache  1@  , Féniès Pierre  2@  , Libo Ren  3@  
1 : LARGEPA
Paris-Panthéon-Assas University
1 rue Guy de La Brosse, 75005 Paris, France -  France
2 : LARGEPA
Paris-Panthéon-Assas University
3 : LARGEPA
Paris-Panthéon-Assas University
1 rue Guy de La Brosse, 75005 Paris, France -  France

In the context of climate change, wildfires are becoming more frequent and intense worldwide, expanding risks to previously unaffected northern regions. One of the most effective strategies for wildfire prevention and early detection is surveillance and intervention patrols. In France, these patrols are primarily conducted by the National Forest Organization relying on daily weather conditions and expert judgment to determine patrol routes. In this work we are interested in a real case study in a department in the South of France. This case study is based on the early detection and prevention strategy implemented by the Departmental Fire and Rescue Service (SDIS) of Gard. During the summer season, the National Forestry Office and the firefighters, using a fleet of vehicles, patrol the forests to prevent fire ignitions. Each day, patrol routes are planned based on tourist presence in forested areas and fire ignition risk levels, which are derived from daily weather reports. Every potential location can be visited at most once within a predefined time-window. The intervention duration varies according to the nature of the operation, whether it involves early fire suppression or population awareness. Every point is assigned a risk level, and locations with higher risk should be prioritized in the routing decisions. The risk level for each point can be interpreted as a score or a profit. The patrol is required to visit certain mandatory locations identified as highly critical, and to take a one-hour lunch break along the route. The patrols start and end in a particular fire station and are subject to a maximum duration of eight hours per day.

In this work, surveillance and intervention patrol is formulated as a team orienteering problem with time windows, mandatory points and a lunch break, where a fleet of vehicles must cover high-risk areas based on tourist activities, fire risk estimations and weather conditions. Patrol routes are restricted by the duration of the daily shift; therefore, not all locations can be visited, and the focus must be on the riskiest ones. Certain locations are designated as mandatory visits due to their high criticality. In addition, the route should include a lunch break stop in a specific time-windows (between 12am and 2pm) . The objective is to maximize the total collected profit (i.e. the total covered risk) of the daily routes.


Chargement... Chargement...